Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Collaborative Uncertainty in Multi-Agent Trajectory Forecasting
Bohan Tang, Yiqi Zhong, Ulrich Neumann, Gang Wang, Siheng Chen, Ya Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add collaborative uncertainty (full covariance) to individual uncertainty | Gaussian noise model on synthetic data · Synthetic Gaussian | KL divergence: 6.68 -> 0.40 | Improved |
| add collaborative uncertainty to individual uncertainty | Laplace noise model on synthetic data · Synthetic Laplace | KL divergence: 12.6 -> 1.65 | Improved |
| add collaborative uncertainty | estimated covariance on synthetic Gaussian · Synthetic Gaussian | l1 of covariance: 1.98 -> 1.01 | Improved |
| add collaborative uncertainty | estimated mean on synthetic Laplace · Synthetic Laplace | l2 of mean: 0.42 -> 0.34 | Improved |
| add Laplace collaborative uncertainty loss to LaneGCN | LaneGCN (our implementation) · Argoverse test | ADE: 1.76 -> 1.70 | Improved |
| add Laplace collaborative uncertainty loss to LaneGCN | LaneGCN (our implementation) · Argoverse test | FDE: 3.84 -> 3.74 | Improved |
| use Gaussian CU loss instead of Laplace CU loss | LaneGCN with CU · Argoverse test | ADE: 1.73 vs 1.70 | Hurt |
| model full covariance instead of diagonal, with interaction module (ablation) | LaneGCN Laplace · Argoverse val | FDE: 3.16 -> 3.11 | Improved |
| model full covariance instead of diagonal, without interaction module (ablation) | LaneGCN Laplace · Argoverse val | ADE: 1.67 -> 1.67 | No clear effect |
| model full covariance instead of diagonal, with interaction module (ablation) | VectorNet Laplace · nuScenes val | FDE: 9.79 -> 9.22 | Improved |
| model full covariance instead of diagonal, without interaction module (ablation) | VectorNet Laplace · nuScenes val | FDE: 10.25 -> 10.21 | Improved |
| model full covariance instead of diagonal, with interaction module (ablation) | LaneGCN Gaussian · nuScenes val | ADE: 4.47 -> 4.39 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- VectorNet — VectorNet: Encoding HD Maps and Agent Dynamics From Vectorized Representation (CVPR 2020) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Unified Uncertainty-Aware Diffusion for Multi-Agent Trajectory Modeling (CVPR 2025 · similarity 0.85)
- Proper Scoring Rules for Agentic Uncertainty Quantification (NeurIPS 2026 · similarity 0.85)
- CUQDS: Conformal Uncertainty Quantification Under Distribution Shift for Trajectory Prediction (AAAI 2025 · similarity 0.84)
- TrajAgent: An LLM-Agent Framework for Trajectory Modeling via Large-and-Small Model Collaboration (NeurIPS 2025 · similarity 0.84)
- HPNet: Dynamic Trajectory Forecasting with Historical Prediction Attention (CVPR 2024 · similarity 0.83)
- C2F-TP: A Coarse-to-Fine Denoising Framework for Uncertainty-Aware Trajectory Prediction (AAAI 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
